telecom optical networks: the road aheadoptical networks: the road ahead biswanath mukherjee child...
TRANSCRIPT
1
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 1
Optical Networks:The Road Ahead
Biswanath MukherjeeChild Family Endowed Chair Professor
University of California, Davis, USA
[email protected]://networks.cs.ucdavis.edu/~mukherje/
Keynote Talk Presented at:4th Euro NGI Conference 2008
Krakow, PolandApril 28, 2008
Telecom
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 2
• This is a reduced / redacted version of the presentation given at Euro NGI 2008.
• For more information on some of the materials, please consult my textbook:
Biswanath Mukherjee, “Optical WDM Networks,”Springer, 2006.
2
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 3
Optical Networks: The Road Ahead
Broadband Access• PONs• WDM in PONs• Long-Reach Broadband Access• Hybrid wireless-optical access• Metro: The vanishing breed?
Backbone Networks• DOCS (“dial for bandwidth”)• Robust Network Design (multi-
layer, multi-domain, multi-path, etc.)
• Ethernet Everywhere-----
• Network Engineering (NE)(vs. TE vs. NP)
• Higher-density switches!
(X)
(X)
Telecom
(X)
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 4
Telecom Network Hierarchy
3
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 5
An Access Network (PON)
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 6
System/Network: Value Proposition
Optical ScienceOptical Science& Engineering& Engineering
CurrentCurrentstate ofstate ofthe artthe art
ElectronicElectronic(Computer)(Computer)EngineeringEngineering
SoftwareSoftwareEngineeringEngineering System
VALUE
System ASystem Ax
x
x
x““MyMy”” systemsystem
++Yours too (?)Yours too (?)
4
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 7
Optical Network Architecture:Extending Our “Boundaries”
Network Architect (Piotr, I, …)
Physical Layer (optical/wireless channel) --materials, devices, subsystems
Applications(“Customer” needs)
+ routing protocols to combat optical channelimpairments
+ breakthroughs needed in device technologies?- optical RAM, ultra-wideband amp, “tunable” AWG, …
Differentiated Services:Bandwidth: OC-192, OC-48, … , STS-1, VT1.5, …Failure-Recovery Delay: The “50-ms myth!”
Network Economics: Pricing, SLA, …
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 8
Broadband Access• PONs• WDM in PONs• Long-Reach Broadband Access• Hybrid wireless-optical access• Metro: The vanishing breed?
Backbone Networks• DOCS (“dial for bandwidth”)• Robust Network Design (multi-
layer, multi-domain, multi-path, etc.)
• Ethernet Everywhere-----
• Network Engineering (NE)(vs. TE vs. NP)
• Higher-density switches!
(X)
(X)
(X)
Optical Networks: The Road AheadTelecom
5
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 9
WDM in PONs (Evolutionary Upgrade)
• Provision Wavelengths as Needed– Distinct growth phases– Upstream/downstream
• Non-Disruptive– Legacy users are unaffected– Infrastructure remains intact
• Maximum Utilization of Capacity– Users are still able to share wavelengths– Shift users requiring more bandwidth
ACK: NSF
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 10
• Shifting (exponential model)• λ0 threshold times: 6, 12.5, 19, 25.5-months
WDM in PONs (Evolutionary Upgrade)
6
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 11
Long-Reach Broadband Access
• Basic Problem– Reach of traditional PON = 10-20 km– Fiber capacity: much larger than data generated by users
served by the limited geography in a traditional PON– Hence Long-Reach (LR) PON should be investigated
• Increase distance to 100 km or more… may need power (so not exactly a “PON”… called “SuperPON” in literature), but use very few active elements... so we call them LR Access Networks
• Sampling of Current LR Access Efforts– ACTS-PLANET; BT Demos; SFI Demos
• Limitations of Current LR Access Methods– “One dimensional” (linear)– Users are located “two dimensionally” on the earth
… so “ring-and-spur” makes more sense for LR Access
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 12
Our LR Access Approach
ACK: ETRI Korea
WR
WR
WR
WR
Metro Ring BackboneNetworkAccess Ring
WRWR
WR
Access Ring
WR
WRWR
WR
Access Ring
WR
ONU
WE-PON
ONU
E/G-PON
WDM-PONONU
P1,1
P1,2
P1,3
P2,1
P2,2
P2,3
P3,1
P3,2
P3,3
100 Km
OLT
ONU
ONU
λ1~ λ16
λ1 ~ λ4λ1
λ2
ONU
λ3 λ4λ5 ~ λ9
λ10 ~ λ16
7
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 13
Hybrid Wireless-Optical Broadband Access Network (WOBAN)
Optimization problem: how far to use fiber? at what point wireless takes over?
ONU
ONU
ONU
Splitter
CO
OLT
AP
Optical backend (PON)Wireless frontend
Wireless routers
OLT
OLT
ONU
Gateway BS
ONU
ONU
ONU
ONU
Splitter
Splitter
Gateway BSGateway BS
Gateway BS
Gateway BS
Gateway BS
Gateway BS
ACK: Nokia (NSN)
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 14
Example: WOBAN Front-End Mesh
A part of San Francisco WOBAN (SFNet)
Location:Golden Gate Ave. to Pacific Ave.Divisadero to Van Ness Avenue
Area: 1 sq-mile (approx.)
Population: 15,000 (approx.)
8
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 15
Wildhorse, Davis, CA: Wireless Routers
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 16
Current Deployment Status
Ack: Suman Sarkar
9
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 17
1 169.237.4.254 (169.237.4.254) 0.973 ms 0.747 ms 0.475 ms
2 169.237.246.238 (169.237.246.238) 0.682 ms 0.797 ms 0.741 ms
3 area2-13--area2.ucdavis.edu (128.120.2.49) 0.880 ms 0.767 ms 0.746 ms
4 area2--area0.ucdavis.edu (128.120.0.133) 0.939 ms 1.295 ms 0.856 ms
5 area0--area14.ucdavis.edu (128.120.0.222) 1.063 ms 1.484 ms 0.887 ms
6 area14--a14.ucdavis.edu (128.120.9.142) 1.009 ms 0.950 ms 0.837 ms
7 a14--ucd-hpr.ucdavis.edu (128.120.9.138) 1.177 ms 1.573 ms 1.173 ms
8 * dc-oak-dc2--ucd-ge.cenic.net (137.164.24.225) 2.901 ms 2.694 ms
9 dc-sfo-dc1--oak-dc2-pos.cenic.net (137.164.22.32) 2.884 ms 2.751 ms 2.655 ms
10 dc-svl-dc1--sfo-dc1-pos.cenic.net (137.164.22.34) 3.923 ms 3.812 ms 3.674 ms
11 te2-3--480.tr01-plalca01.transitrail.net (137.164.131.253) 3.854 ms 4.300 ms 3.958 ms
12 bb1-g1-0.pxpaca.sbcglobal.net (198.32.176.112) 4.543 ms 4.463 ms 4.559 ms
13 151.164.93.249 (151.164.93.249) 68.968 ms 68.869 ms 69.440 ms
14 dist1-vlan30.lgtpmi.ameritech.net (65.42.245.97) 69.434 ms 69.213 ms 68.944 ms
15 rback1-g1-0.lgtpmi.sbcglobal.net (65.42.245.230) 68.845 ms 69.043 ms 69.145 ms
16 adsl-68-22-232-254.dsl.lgtpmi.ameritech.net (68.22.232.254) 823.731 ms 856.567 ms 822.425 ms
17 adsl-68-22-232-249.dsl.lgtpmi.ameritech.net (68.22.232.249) 858.838 ms 890.727 ms 859.053 ms
Traceroute: to aland.bbn.com
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 18
DOCS Architecture: Overview
ACK: NSF FIND Program, Dan Blumenthal, Nick McKeown, and John Bowers
10
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 19
Why DOCS? (or DCS)
• Emerging (video-enabled) applications:– Video downloads– Massively multiplayer games– Video collaborations– Telepresence– IPTV– Applications on a wire, etc.
• If you are happy with the PMO of our networks:– slow downloads– jittery streaming– unreliable audiothen DOCS is not for you.
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 20
DOCS Architecture – Physical Aspects
• Circuit switches are suitable to optics (McKeown et al.)– Require no packet processing or buffering
• Circuit switches are simple– Requires about 90% less hardware than an equivalent packet switch– More attractive than equivalent packet switches:
• About eight times the capacity of a packet switch• Five times the capacity per watt• Cost one quarter of the price of the equivalent packet switch.
– Simplest way to build a circuit switch is to “Start with a packet switch and throw most of it away” [Pablo-HotNets-I 2002, IEEE MICRO 2002, JON 2003]
• Circuit switches perform well– Fast restoration times (less than 50 ms)– Contain less software than equivalent routers– Good service guarantees: no queuing delay (so predictable packet
delays), and guaranteed bandwidth
11
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 21
DOCS: Basic Premise
• DOCS = Dynamic Optical Circuit Switching
• Approach:– Bursty (packet) traffic generated by users/applications– Aggregate traffic at the network edge– Establish high-bandwidth pipes between edge nodes
through the network core– DOCS offers bandwidth-on-demand capabilities to
applications (users can “dial” for bandwidth)
• Example Applications:– Real-time download (say within 5 sec)– Database/website backup (say between 1 am – 3 am,
and not to exceed a 15-min duration)
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 22
New Trends for Future Telecom Backbone Networks
• Emergence of new bandwidth-hungry applications:– Scientific grid-computing applications– Consumer applications: Video-on-Demand (VoD), IPTV, …
• Emerging services that require dynamism and flexibility:– Flexible bandwidth reservation, e.g., Bandwidth on Demand
(dial for bandwidth), Advanced reservations, …– Service guarantees (guaranteed bandwidth, fast switching/
restoration times, SLA, availability)
• Increasing gap between optical and electronic transmission and processing speeds
12
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 23
DOCS Example Applications - DDRs
• DDR = Deadline-Driven Request• Real-time file download
– maximum transfer time a person can tolerate (perhaps 5 sec)
• Database/website/server backup: scheduled at any time, perhaps during night, but with a fixed deadline (completion time)
– flexible application requiring high bandwidth, but not necessarily instantaneously.
• Our research: Provision DDRs in a DOCS network from two perspectives:
– Time dimension: allows flexibility w.r.t. when to when to schedule the request.
– Transmission bandwidth (rate) dimension: allows flexibility (adjustable to network state) w.r.t. data rate to be allocated to the request.
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 24
Ethernet is a success story in Local Area Networks (LAN)About 90% of LANs use Ethernet.
Extending its reach from LAN into Metro Area Networks (MAN) has already been established.
Focus now is to extend Ethernet into carrier core networks.Future mode of operation: Ethernet over WDM native Ethernet frames directly over WDM.
Elimination of several layers of other technologies.CapEx and OpEX savings.
Connection-oriented Ethernet.Forwarding: VLAN-XC, Provider Backbone Transport (PBT), T-MPLS.
Following requirements must be taken into account: High resilience.Long reach: 1500- 4000 km.Rates of up to 100 Gbit/s Ethernet (GbE).High degree of mesh.
Ethernet Everywhere
ACK: Siemens (NSN)
13
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 25
Transmission Rates
Ethernet Rates:100 Gbit/s Ethernet.
Max possible CapEx savings.
Constraint: Signal transmission range for a certain rateSignal’s quality depends on the physical impairments.Transmission Range = Signal traveled distance after which signalquality degrades to a level that it needs regeneration.
Transmission Ranges:Range of 10 Gbit/s signal = 3000 Km.Range of 100 Gbit/s signal = 500 Km.
Etherpath = Lightpath carrying Ethernet frames.
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 26
Telecom Nets: “End-to-End” Ethernet?
MetroMetro Long-Haul (Backbone)Long-Haul (Backbone) MetroMetro
Residential
EnterpriseHQ
ISP
Access Technologies
Branch
Branch
VoD/HDTV
Data Center
InternetInternet
14
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 27
TE vs. NE vs. NP
• Traffic Engineering (TE)
• Network Engineering (NE)
• Network Planning (NP)
• TE – online, dynamic, provisioning problem, ms time scale
• NE – intermediate problem, months time scale
• NP – offline, static, dimensioning problem, 5-yr time scale
– “Put the traffic where the bandwidth is”
– “Put the bandwidth where the traffic is”
– “Put the bandwidth where the traffic is forecasted to be”
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 28
Robust Network Design
TTR
0 tTime
Fault occurs Fault fixedConnection arrival
……
Connection departure
……Network Events:
Network Architecture Needs:• More dynamism / flexibility / agility / automation …• Handle multiple (near-simultaneous) faults efficiently• Reprovision (backup) capacity when “network state” changes
– Efficiency tradeoff: bandwidth vs. implementation
• Is a service path “ahead” or “behind” the contract (SLA)?– Reprovision, if necessary; role of penalty and economics in SLA
• Utilize “excess capacity”?• …
15
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 29
Further Reading
Broadband Access• PONs• WDM in PONs – IEEE Photonics in Switching Conf. (PIS) , Aug. 07• Long-Reach Broadband Access – stay tuned• Hybrid wireless-optical access – JLT, Nov. 07; OFC’07; ICC’07, …• Metro: The vanishing breed?
Backbone Networks• DOCS (“dial for bandwidth”) – PIS, Aug. 07• Robust Network Design – Lots of literature• Ethernet Everywhere – OFC’07 PD paper; GB’07; JLT, Jan. 08
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 30
Extra Slides
16
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 31
What Is An Optical Network?
• It is NOT NECESSARILY all optical” ” ” ” packet switched
• Characteristics of an optical network– Transmission: optical– Switching: could be optical, could be electronic, could be hybrid
could be circuit, could be packet, could be burst
• Most Promising Approach Today– Electronic circuit switching with sub-lambda granularity (STS-1, STS-3, …)
• Example Utility for IP Networking– Connect any two IP routers (geographically far apart) with a direct
(“virtual”) bandwidth pipe… of whatever capacity (STS-1, … , STS-192)– Increase (or decrease or delete) the capacity on demand– Dynamically control the “topology” connecting the IP routers– Create a “separated control network” (of whatever bandwidth)– …
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 32
Resilient Mesh Net --Overview
s di
Working
Backup
• Multi-path routing– How many paths?– State-dependent path capacities– Virtual concatenation (VC)
(“traffic grooming”)
1sdp
bsdp
3sdp
2sdp
• Survivable Routing– Primary + backup path(s)– Precomputed backup (?)– Shared backup (?)– Need reliable path (five-9’s)
• Component/link availability
– Need fast recovery
17
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 33
Data over (Next-Gen) SONET/SDH
Ethernetport GFP VC
LCASEthernet
portGFPVCLCAS
SONET/SDH-enabled optical transport
network
GbE GbE
SONET/SDHframer/mapper
SONET/SDHframer/mapper
STS-3c-7v
7 STS-3c’s
STS-3c-7vSource Destination
• Inverse multiplexing• Relaxing time-slot contiguity/alignment• Reassembly and jitter control
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 34
Next-Gen SONET Chip Availability
Source: Light reading
18
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 35
Problem StatementProblem Statement
• Given:– Network topology G (V, E, C(e, w))– Basic time-slot component, e.g., VT1.5, STS-1, STM-1, etc.– A high-speed request R with bandwidth requirement B. – An inverse-multiplexing control parameter K for maximal
allowed path number.
• Find:
– m distinct paths, where 1 ≤ m ≤ K, such that the aggregated capacity they offer ≥ B.
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 36
Implication to Service Resilience
• Route splitting enables degraded services– Ethernet service: peak rate 600 Mbps, guaranteed rate
300 Mbps (against single network-element failure)– Need LCAS
Source
Destination
LCAS
300 Mbps
300 Mbps
19
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 37
Motivation:Critical Infrastructure Survivability
• Overloads, Attacks, and Failures:– What’s the difference? It’s a “Race against Time”
• “Our society depends on critical infrastructures for delivery of essential services.
• “Telecom, banking, power supply, public transport, etc. increasingly rely on information infrastructures… not only for management and control… but also for monitoring outages and recovery.
• “Combinations of wireless, ad hoc, and fixed networks are becoming a reality in many domains.”
• Tomorrow's networks face the survivability challenge:– how to deliver critical services in a timely manner in
presence of overloads, attacks, and failures?
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 38
Modeling “Disasters”
• Define SRGs (shared-risk groups)– for (potentially) correlated failing components– e.g., links + nodes in close geographical proximity
• SRLG (shared-risk link group)– Special case of SRG– Quite common since fibers are laid in ducts, and
duct cuts are more frequent than we wish
• Extend link models (or develop new models) to apply to SRGs
20
April 28, 2008 © B. Mukherjee, UC Davis: “Optical Nets: The Road Ahead” (EuroNGI-08) 39
Which Method?
• PREV: Provisioning fast REstorable VCG– One backup path per node
pair
s di
Working
Backup
1sdp
bsdp
3sdp
2sdp
• PIVM: Protecting Individual VCG Member– One working VCG per
connection– One backup VCG per
working VCG member
s d
1wp
1,1bp
2wp
2,2bp
1,2bp
i
WorkingBackup